Realtime Player's Posture Measurement System for Air-Hockey Robot

Abstract

In this research, a real-time estimation system of an opponent's posture in an air-hockey robot system. A posture of an opponent player's is estimated at a rate of 500 Hz by integrating a 3D depth sensor and a high-speed vision system through extended Kalman filter. In addition, we propose a new method to decide an attack direction based on manipulability so as to make it difficult for a player to hit back a puck. Experimental results show the effectiveness of the proposed method compared with a previous method.